ReframeYour lawyers already work across several AI tools
Your lawyers draft in Copilot, research in Claude and test arguments in ChatGPT. Each tool sees only the files a lawyer pastes into the chat. None of those tools sees the ethical walls that screen a matter.
At one Am Law 100 firm, the AI engineers connected iManage, permissions and matter context to each tool by hand. The engineers repeated that work for every new tool and enforced the walls again in every connection.
One governed source, read in place
Reframe reads your documents where they are, in iManage, SharePoint and Microsoft 365, and moves no documents. Reframe reads each document against your legal ontology: the clients, matters, funds, agreements, provisions and dates a practice works with, and how they relate.
Every fact links to the page it came from. Reframe refuses a fact that contradicts the ontology and saves the reason for review.
Read what a legal ontology is and how a legal knowledge graph works for a law firm.
The same cited answer in every AI tool
Claude, ChatGPT, Microsoft 365 Copilot and the agents your firm builds connect to Reframe over the Model Context Protocol (MCP). Each tool signs in as the lawyer who asks.
Reframe checks the lawyer's walls and permissions before the tool reads a single passage. Reframe logs every question, answer and refusal in an append-only audit log.
Ask Claude "What leverage ratio applies to Project Grove?" and the answer cites the credit agreement, page 38, and Amendment No. 2, page 4. Ask Copilot the same question, and you get the same answer with the same pages. Ask about a matter you are screened from, and Reframe refuses the request and logs the refusal.
Measured on 1,000 fund documents
Reframe tested the same AI model on fund formation questions, with and without the Reframe graph. With Reframe, the model scored 99.5% accuracy on fund terms across 66 funds the system had never seen.
Cost per criterion met improved 85% against an agent reading raw files. On a set of eight questions, the model answered 8 of 8 correctly with the graph and 0 of 8 reading raw files.
Three ways firms try to get there
An in-house build costs millions and a permanent engineering team. A vendor platform improves the vendor's product with your work product. A data lake takes years before a lawyer sees an answer.
Reframe starts with one practice, and your people run the platform when the program ends.
How your firm starts
Start with a 90-minute working session on the questions lawyers in one practice ask. Reframe moves no documents for the session.
Reframe and your knowledge team then define the ontology of that practice, connect the sources and validate the graph against your documents. Forward-deployed Reframe engineers work inside your firm until your people run the platform, and your firm keeps the ontology and every deliverable.